Optimal surveillance intervals for hepatocellular carcinoma screening in cirrhotic patients with hepatitis C infection: a Taiwanese national cohort study

Abstract

Background

Hepatocellular carcinoma (HCC) surveillance guidelines recommend ultrasound screening every 6 months, which was determined empirically. Since liver disease progression is heterogeneous among different etiologies, it is clinically valuable to analyze patients with specific etiologies. The aim of this study was to evaluate the impact of surveillance intervals, duration of cirrhosis, and HCC risk factors on the survival of hepatitis C virus (HCV) cirrhotic patients.

Methods

This nationwide cohort study included adult patients who were newly diagnosed as having HCV-related cirrhosis between January 2007 and December 2018. 5,093 newly diagnosed cirrhotic HCV related HCC patients were analyzed. The timing of ultrasonography screening was categorized into 4 cohorts: 0- to 6-month cohort (6-month cohort), 7- to 12-month cohort (12-month cohort), 13- to 24-month cohort (24-month cohort), and not screened within 2 years cohort (unscreened cohort). The chance of early stage of HCC diagnosis and receiving curative treatment were calculated. Association between surveillance interval and all-cause mortality was analyzed adjusting for lead-time bias.

Results

The 6-month group had the highest likelihood of being diagnosed with an early-stage HCC, followed by the 12-month group (OR = 0.69; 95% CI 0.55–0.85) and the 24-month group (OR = 0.355; 95% CI 0.27–0.47), the last by the unscreened group (OR = 0.296; 95% CI 0.22–0.40). The 6-month group had the highest likelihood of being received curative treatment, followed by the 12-month group (OR = 0.721, 95% CI 0.58–0.89) and the 24-month group (OR = 0.584; 95% CI 0.44–0.77), the last by the unscreened group (OR = 0.513; 95% CI 0.38–0.69). The 6-month group had the least likelihood of all-cause mortality, followed by the 12-month group (HR = 1.134; 95% CI 1.02–1.26), the 24-month group (HR = 1.570; 95% CI 1.39–1.77), and the unscreened group (HR = 1.520; 95% CI 1.33–1.73). After adjusting for lead-time bias, the 6-month group had the least likelihood of all-cause mortality. In the 6-month group, cirrhotic HCV patients with an AFP less than 20 ng/ml, with a MELD score less than 20, with cirrhosis duration between 3–5 years had better survival.

Conclusion

A 6-month surveillance interval can significantly improve the detection rate of early-stage HCC, likelihood of receiving curative treatment, and prolong the overall survival of cirrhotic HCV patients.

Introduction

Hepatocellular carcinoma (HCC) remains a leading cause of cancer-related death worldwide [1]. HCC mortality rate increased between 2000–2013, plateaued during 2013–2016, and began to decline during 2016–2018 in USA [2]. A series of Taiwan nationwide interventions have been implemented for HCC prevention [3], but HCC remains a critical public health problem. One most important factor determining the effectiveness of a surveillance program is the length of surveillance interval. Therefore, despite differences in health policies, insurance, screening, and medical costs, international guidelines strongly recommend regular HCC surveillance with ultrasound every six months in patients with cirrhosis; rather than annually [4–8].

Improvements in HCC survival in the past decades have been attributed to increased early detection and development in HCC treatment [9]. The median of 5-year survival following the curative treatment among patients with Barcelona Clinic Liver Cancer (BCLC) stage A ranges between 50% −70% [4]. Earlier retrospective analyses which took lead time bias into consideration concluded that HCC surveillance was an independent predictor of survival [10–13]. But some systemic research in the past concluded that the evidence supporting surveillance of high-risk populations in reducing HCC mortality had very-low-strength [14, 15].

The HCC mortality rate in Taiwan follows the same trend as that in the United States [2], with a slight decrease in recent years. However, the prevalence and incidence of HCC remain high [2, 3]. Previous surveillance interval studies have not analyzed whether survival benefits of surveillance intervals differ by HCC etiologies [10, 11]. It is likely that the survival rate and burden of HCC vary by etiology. For example, patients with alcohol related diseases often had poor compliance compared with other etiologies [16]. Moreover, hepatitis B virus (HBV)-related HCC has a better survival rate in the United States [17]. Only few studies focus on hepatitis C virus (HCV)-related cirrhosis and the evidences are limited to the impact of surveillance interval on overall HCC [18]. The understanding in the Asia–Pacific region is far limited. In this study, in order to more precisely characterize HCC-etiology specific surveillance intervals, we focused on patients with HCV-related cirrhosis.

The current surveillance guidelines were determined empirically [4–6] and were not validated through well-designed randomized trials. AASLD and EASL Guidelines recommend abdominal ultrasound every 6 months to detect early stage of HCC, thereby improving survival [4, 5]. The six-month interval is currently a reasonable choice because past studies have shown that a shorter interval like three-month interval does not provide any clinical benefit and ultimately leads to unnecessary surveillance and medical costs [4, 19]. According to the literature, commonly used intervals were 6, 12, and 24 months [10–12]. The time intervals used in previous literature included 6 months (less than 7 months) [12], 6, 12, 24 months [11], and 24 months [10]. A well-designed randomized controlled trial comparing surveillance versus no surveillance or a 6-month versus 12-month time frame would be ideal for resolving controversies, but it may be impractical or unethical to conduct such a trial [12, 20]. Currently, there were three prospective cohort studies using hospital data, but their sample size and representativeness were somewhat limited [12, 21, 22].

The aim of this study was to evaluate the impact of ultrasound surveillance intervals, duration of HCV-related cirrhosis, and potential risk factors for HCC on survival. Therefore, we used the Taiwan Health Database to determine which characteristics or subgroups in patients with HCV-related cirrhosis could achieve better survival if they had shorter intervals like semi-annual HCC surveillance compared with annual and biannual surveillance [10–12].

Methods

Data source

The primary data source is the 2007–2018 National Health Insurance Research Database (NHIRD), which is managed by the National Health Research Institutes and available for research purposes with a proper application and review process. Taiwan’s NHI program provides universal health insurance to all residents in Taiwan (approximately 23 million people). We used specific files in the NHIRD, such as enrollment files and facility registries, in this study. The enrollment files provide information on each person’s monthly wage and category, location, and date and reason for unenrollment. The NHI hospital registry provides information on the accreditation level of each physician’s practice setting. This cohort study included HCC patients from the Taiwan Cancer Registry, which has comprehensive and high-quality data on cancer patients. Patient identification numbers are all encrypted in the database.

Study population and cohort

This nationwide retrospective cohort study included the data of adult patients who were newly diagnosed as having HCV-related cirrhosis between January 2007 and December 2018. We defined the index date of screening for HCC as 90 days before the date of HCC diagnosis [11]. The timing of ultrasonography screening prior to the index date was used to categorize patients into 4 cohorts: 0- to 6-month cohort (6-month cohort), 7- to 12-month cohort (12-month cohort), 13- to 24-month cohort (24-month cohort), and not screened within 2 years cohort (unscreened cohort). We included the data of only patients with an HCV cirrhosis diagnosis before the index date. We excluded patients aged younger than 20 years and patients who had a history of chronic hepatitis B (CHB), human immunodeficiency virus (HIV) infection, solid tumor, and hematologic malignancy. We also excluded received liver surgery before index date. The total numbers of ultrasonography in the 3 years prior to HCC diagnosis was also analyzed. Comorbidities analyzed in the current study included diabetes mellitus (DM), mental illness including schizophrenia, depression, anxiety, and dementia, coronary artery disease (CAD), cerebrovascular disease (CVD), chronic obstructive pulmonary disease (COPD), end stage renal disease (ESRD), and alcohol-related liver diseases (ALD).

HCC diagnosis and staging

HCC stage at diagnosis was classified using the BCLC system. Patients with 0/A early-stage disease were analyzed and compared with those with advanced disease at higher BCLC stages.

Chance to receive curative treatment

HCC therapeutic methods are significantly associated with overall HCC mortality. We defined patients who underwent liver resection, liver transplantation or ablation within 1 year after HCC diagnosis as those who received curative treatment. Patients who received trans-arterial chemoembolization (TACE), sorafenib, radiotherapy or no treatment within 1 year after HCC diagnosis were defined as patients with non-curative treatment.

All-cause mortality

All-cause mortality for all HCC patients was measured from the index date to the date of death or the end of 2018. Death was defined as withdrawal from the compulsory NHI program.

Patient characteristics

Patient age, sex, preexisting comorbidities, health-care-seeking behavior, and socioeconomic status (SES) were analyzed. The comorbidities defined in this study were DM, mental illness, CAD, CVD, COPD, ESRD, ALD within the 2 years preceding the index date. Patients with preexisting DM, mental illness, CAD, CVD, or COPD were defined as those with one inpatient diagnosis of one of the aforementioned diseases or with at least three outpatient visits for one of the aforementioned diseases and disease-related prescriptions. Patients diagnosed with ESRD were defined as those enrolled in the Catastrophic Illness Patient Registry. Patients with preexisting ALD was defined as those with one inpatient diagnosis of one of the aforementioned diseases or with at least three outpatient visits for one of the aforementioned diseases. Outpatient utilization was based on the number of outpatient visits made by a patient within the 2 years preceding the index date; it was categorized into three levels: low, intermediate, and high. Furthermore, we used the NHI monthly wage/category as a proxy of SES. Patients with a well-defined monthly payroll were categorized into four groups according to their monthly payroll level: ≥ NT$40,000, NT$39,999–NT$25,000, NT$15,000–NT$24,999, and < NT$14,999 (NT$: New Taiwan dollars; US$1 = approximately NT$30.24). We also categorized the urbanization level into three groups, namely, urban, suburban, and rural. Certain drugs that may alter the risk of HCC, such as aspirin, non-steroidal anti-inflammatory drugs (NSAIDs), cyclooxygenase-2- specific inhibitors, metformin, and statin were analyzed. Use of these medications was defined as use more than 10% of the time in the 2 years before the index date. We also analyzed the impact of tumor potential risk factors at HCC diagnosis, such as Alpha-Fetoprotein (AFP), Child–Pugh score, and Model for End-Stage Liver Disease (MELD) score, on survival rate.

Provider characteristics

We considered the patient's principal health care provider to be the patient's primary provider of ultrasound examinations within the 3 years before the index date. Under the NHI program in Taiwan, medical institutions are classified according to their quality, staffing, and infrastructure. Accordingly, accreditation level was divided into two categories: (a) medical center and (b) non-medical center. Hospital ownership was considered either public or private.

Statistical analysis

We compared the demographic characteristics of the four groups of HCC patients using the χ test for dichotomous or nominal variables and the Kruskal–Wallis test for continuous variables. Odds ratios (ORs) calculated using logistic regression were used to investigate the association between surveillance interval and early HCC diagnosis and curative treatment. Additionally, Cox proportional hazards regression (HR) was performed to analyze the association between surveillance intervals and all-cause mortality. After univariate analysis of various demographic and clinical factors associated with early diagnosis of HCC, curative treatment of HCC, and all-cause mortality, multivariate analysis was further performed using logistic regression and Cox regression. To adjust for potential lead-time bias, we used the parametric model proposed by Duffy et al. [23], assuming an exponential distribution of the HCC sojourn time. The HCC sojourn time was assumed to be 70, 140, and 220 days, according to the estimations reported in previous studies [11, 24]. Survival rates were estimated using a Kaplan–Meier approach. It is difficult to confirm whether the missing data in the Taiwan Cancer Registry are missing at random (MAR). We did not perform multiple imputation techniques to handle missing data. Instead, we conducted sensitivity analyses by excluding individuals with missing values. All statistical analyses were conducted using SAS version 9.4 software (SAS Institute, Cary, NC, USA). A p value less than 0.05 was considered statistically significant.

Ethical approval

The study protocol was approved by the Taipei City Hospital Research Ethics Committee (TCHIRB-11110004-E).

Results

Characteristics of study patients

Between 1 January 2010 and 31 December 2018, a total of 79,112 HCC patients were identified. Only the 20,263 patients who also had HCV infection before the index date were enrolled. We then excluded 4,785 patients with HBV infection, 7,760 patients without cirrhosis before and the index date, 1,042 patients with solid cancer, 52 patients with hematological malignancy, 263 patients with liver surgery or transplantation before the index date, 20 patients with HIV, 1224 HCC patients were diagnosed between April 2017 and December 2018, and 24 patients with a data missing. Finally, 5,093 newly diagnosed cirrhotic HCV related HCC patients were finally analyzed. Among them, we defined the 3,862 individuals who underwent ultrasound surveillance 0–6 months before the index date as the 6-month group. 520 and 346 patients underwent ultrasound surveillance for 7–12 months and 13–24 months, and 365 patients did not undergo ultrasound surveillance within 2 years and were classified as the 12-month group, the 24-month group, and the never screened group, respectively (Fig. 1).

Fig. 1.

Flowchart of patient selection

Distribution of patient and provider characteristics of the study cohort by surveillance groups were as shown in Table 1. The percentages of female patients were highest in the 6-month group (53.03%). The percentages of elderly aged 75 years or above were lower in the 6-month group (29.98%). The prevalence of current alcohol drinking (7.48%) and current smoking (14.27%) were lower in the 6-month group. The percentages of patients with a high-level of outpatient utilization were highest in the 6 M group (34,46%). The percentages of people with BCLC stage 0 (10.88%) and stage 1 (41.74%) were higher in the 6-month group. Other baseline characteristics were detailed in Table 1.

Table 1.

Baseline demographic characteristics of the study cohort

Categorial Variables	6-month Group		12-month Group		24-month Group		Unscreened group		P
	n = 3862		n = 520		n = 346		n = 365
	N	%	N	%	N	%	N	%
Gender									0.00
Male	1814	46.97	257	49.42	182	52.60	206	56.44
Female	2048	53.03	263	50.58	164	47.40	159	43.56
Age									0.00
Age < 55	322	8.34	36	6.92	30	8.67	29	7.95
55 ≦ Age < 65	1023	26.49	112	21.54	83	23.99	86	23.56
65 ≦ Age < 75	1359	35.19	183	35.19	99	28.61	114	31.23
Age ≥ 75	1158	29.98	189	36.35	134	38.73	136	37.26
Alcohol drinking									0.10
Never	1908	49.40	260	50.00	150	43.35	164	44.93
Past	335	8.67	34	6.54	33	9.54	29	7.95
Current	289	7.48	40	7.69	31	8.96	41	11.23
Missing	1330	34.44	186	35.77	132	38.15	131	35.89
Smoking									0.0908
Never	1821	47.15	237	45.58	141	40.75	155	42.47
Past	88	2.28	12	2.31	16	4.62	9	2.47
Current	551	14.27	82	15.77	50	14.45	62	16.99
Missing	1402	36.30	189	36.35	139	40.17	139	38.08
Co-morbidities
DM	773	20.02	106	20.38	62	17.92	31	8.49	<.0001
Mental illness	61	1.58	11	2.12	8	2.31	5	1.37	0.5953
CVD	118	3.06	22	4.23	21	6.07	9	2.47	0.0112
COPD	127	3.29	26	5.00	21	6.07	7	1.92	0.0042
ESRD	134	3.47	27	5.19	21	6.07	7	1.92	0.0059
ALD	188	4.87	19	3.65	19	5.49	6	1.64	0.0218
Medications
Metformin	745	19.29	86	16.54	58	16.76	51	13.97	0.0352
Statin	262	6.78	30	5.77	27	7.80	15	4.11	0.1513
Aspirin	368	9.53	46	8.85	53	15.32	46	12.60	0.0016
Seeking behavior									<.0001
Low	1162	30.09	166	31.92	153	44.22	214	58.63
Intermediate	1369	35.45	183	35.19	93	26.88	85	23.29
High	1331	34.46	171	32.88	100	28.90	66	18.08
SES									0.0545
0–14,999	715	18.51	109	20.96	79	22.83	89	24.38
15,000–24,999	2133	55.23	286	55.00	187	54.05	195	53.42
25,000–39,999	456	11.81	57	10.96	40	11.56	28	7.67
≥ 40,000	558	14.45	68	13.08	40	11.56	53	14.52
Urbanization									0.8714
Urban	552	14.29	73	14.04	51	14.74	47	12.88
Suburban	2086	54.01	270	51.92	192	55.49	199	54.52
Rural	1224	31.69	177	34.04	103	29.77	119	32.60
Hospital Level									<.0001
Not center	2592	67.12	358	68.85	248	71.68	223	61.10
Center	1270	32.88	162	31.15	98	28.32	92	25.21
No sonography	0	0.00	0	0.00	0	0.00	50	13.70
Hospital type									<.0001
Private	3018	78.15	405	77.88	276	79.77	244	66.85
Public	844	21.85	115	22.12	70	20.23	71	19.45
No sonography	0	0.00	0	0.00	0	0.00	50	13.70
BCLC Stage									<.0001
Stage 0	420	10.88	42	8.08	9	2.60	9	2.47
Stage 1	1612	41.74	186	35.77	90	26.01	82	22.47
Stage 2	494	12.79	79	15.19	65	18.79	62	16.99
Stage 3	408	10.56	75	14.42	70	20.23	83	22.74
Stage 4	161	4.17	27	5.19	35	10.12	33	9.04
Missing	767	19.86	111	21.35	77	22.25	96	26.30
Tumor Size									<.0001
≦ 1 cm	88	2.28	16	3.08	4	1.16	5	1.37
1 cm < T ≦ 3 cm	2197	56.89	229	44.04	111	32.08	105	28.77
3 cm < T ≦ 5 cm	738	19.11	115	22.12	85	24.57	76	20.82
> 5 cm	319	8.26	75	14.42	92	26.59	111	30.41
Missing	520	13.46	85	16.35	54	15.61	68	18.63
Treatment
Surgery	390	10.10	50	9.62	35	10.12	24	6.58	0.1916
Ablation	1007	26.07	96	18.46	42	12.14	48	13.15	<.0001
TACE	1581	40.94	220	42.31	145	41.91	142	38.90	0.7639
Others	884	22.89	154	29.61	124	35.83	151	41.37	0.184
AFP (ng/mL)									<.0001
< 10	633	16.39	70	13.46	46	13.29	45	12.33
10–20	371	9.61	52	10.00	19	5.49	23	6.30
21–400	1001	25.92	123	23.65	62	17.92	68	18.63
> 400	339	8.78	60	11.54	63	18.21	59	16.16
Missing	1518	39.31	215	41.35	156	45.09	170	46.58
Child Score									<.0001
A	1705	44.15	206	39.62	111	32.08	102	27.95
B	475	12.30	64	12.31	47	13.58	71	19.45
C	98	2.54	20	3.85	21	6.07	19	5.21
Missing	1584	41.02	230	44.23	167	48.27	173	47.40
MELD Score									<.0001
< 10	1449	37.52	182	35.00	101	29.19	103	28.22
10–20	594	15.38	86	16.54	55	15.90	57	15.62
> 20	666	17.24	100	19.23	89	25.72	104	28.49
Missing	1153	29.85	152	29.23	101	29.19	101	27.67
Follow up time	1040	763	920	743	715	665	713	659	<.0001
All mortality	2670	69.14	384	73.85	291	84.10	298	81.64	<.0001
HCC diagnosis									0.1956
2010	481	12.45	59	11.35	33	9.54	26	7.12
2011	539	13.96	78	15.00	41	11.85	43	11.78
2012	533	13.80	79	15.19	49	14.16	47	12.88
2013	571	14.79	79	15.19	55	15.90	48	13.15
2014	546	14.14	63	12.12	50	14.45	59	16.16
2015	506	13.10	70	13.46	44	12.72	61	16.71
2016	547	14.16	74	14.23	61	17.63	64	17.53
2017	139	3.60	18	3.46	13	3.76	17	4.66
Sonography No	7.90	3.62	4.87	2.61	3.17	1.88	1.16	0.82	<.0001
Cirrhosis duration									<.0001
< 1 year	579	14.99	66	12.69	6	1.73	27	7.40
< 3 years (≥ 1 year)	928	24.03	108	20.77	119	34.39	62	16.99
3–5 years	1197	30.99	179	34.42	94	27.17	143	39.18
> 5 years	1158	29.98	167	32.12	127	36.71	133	36.44

DM Diabetes mellitus, Mental illness including schizophrenia, depression, anxiety, and dementia, CVD Cardiovascular disease, COPD Chronic obstructive pulmonary disease, ESRD End-stage of renal disease, ALD Alcohol-related liver disease, Surgery Included liver re section and transplantation, TACE Trans-arterial chemoembolization, Ablation Included radiofrequency ablation, and percutaneous ethanol injection therapy, Others Included systemic treatment and no treatment, Aspirin including aspirin, non-steroidal anti-inflammatory drugs, and Cyclooxygenase-2 inhibitors, SES Socioeconomic status: 0–14,999(NTD$), 15,000–24,999 (NTD$), 25,000–39,999 (NTD$), > 40,000 (NTD$), Center: medical center (tertiary hospital), No sonography no sonography examination, AFP Alpha-fetoprotein, Child Score Child–Pugh score, MELD Score Model for End-Stage Liver Disease score, Follow up time days of follow-up duration in the study, All mortality All-cause mortality, Sonography No Numbers of sonography examination

Chance of early stage of HCC diagnosis

The proportions of patients in the 6-month, 12-month, 24-month and unscreened groups who were diagnosed as at the BCLC stage 0 or 1 were 65.65%, 55.75%, 36.80% and 33.83%, respectively. The univariate analyses showed that a clear dose response pattern was observed between surveillance interval and early-stage HCC diagnosis. The 6-month group had the highest likelihood of being diagnosed with an early-stage HCC diagnosis, followed by the 12-month group (OR = 0.659; 95% CI 0.54–0.81) and the 24-month group (OR = 0.31; 95% CI 0.24–0.40), the last by the unscreened group (OR = 0.27; 95% CI 0.21–0.35). Likewise, female sex, patients with aged 55–64 or aged 65–74, patients with intermediate or high outpatient utilization, and patients with a monthly wage 15,000–24999 NTD, 25,000–39,999 NTD, or ≥ 40,000 NTD were more likely to be diagnosed with an early stage of HCC. In contrast, patients with past or current alcohol drinking behavior, patients who currently smoke, patients with CVD, ESRD, and ALD, and patients with metformin use, statin use, or aspirin use were significantly more likely to be associated with late stage of HCC diagnosis (Table 2).

Table 2.

Univariable and multivariable analyses of variables associated with early stage of HCC diagnosis

	Event/Patients/%	Univariate			Multivariable
		OR	95%	P	OR	95%	P
Screening interval
0–6 months (ref)	2032/3095/65.65
7–12 months	228/409/55.75	0.659	(0.54, 0.81)	<.0001	0.686	(0.55, 0.85)	0.0007
13–24 months	99/269/36.80	0.305	(0.24, 0.40)	<.0001	0.355	(0.27, 0.47)	<.0001
Unscreened	91/26.9/33.83	0.267	(0.21, 0.35)	<.0001	0.296	(0.22, 0.40)	<.0001
Gender
Male (ref)	1109/1939/57.19
Female	1341/2103/63.77	1.317	(1.16, 1.50)	<.0001	1.144	(0.98, 1.34)	0.0934
Age
Age < 55	193/345/55.94
55 ≦ Age < 65	682/1074/63.50	1.370	(1.07, 1.75)	0.0122	1.082	(0.82, 1.42)	0.5713
65 ≦ Age < 75	893/1416/63.06	1.345	(1.06, 1.71)	0.0149	1.037	(0.79, 1.37)	0.7987
Age ≥ 75	682/1207/56.50	1.023	(0.80, 1.30)	0.8528	0.920	(0.69, 1.23)	0.573
Alcohol drinking
Never	1516/2375/63.83
Past	229/416/55.05	0.694	(0.56, 0.86)	0.0007	0.747	(0.57, 0.98)	0.0346
Current	186/377/49.34	0.552	(0.44, 0.69)	<.0001	0.646	(0.49, 0.85)	0.0018
Missing	519/874/59.38	0.828	(0.71, 0.97)	0.0202	1.074	(0.60, 1.91)	0.8072
Smoking
Never	1422/2522/63.14
Past	66/121/54.55	0.700	(0.49, 1.01)	0.0579	0.953	(0.62, 1.47)	0.8258
Current	390/713/54.70	0.705	(0.59, 0.84)	<.0001	1.043	(0.82, 1.33)	0.728
Missing	572/956/59.83	0.869	(0.75, 1.02)	0.077	1.626	(1.00, 2.66)	0.0519
Co-morbidities
DM	436/758/57.52	0.854	(0.73, 1.00)	0.0532	0.964	(0.80, 1.17)	0.7103
Mental illness	42/72/58.33	0.907	(0.57, 1.46)	0.6869	1.007	(0.60, 1.68)	0.979
CVD	61/124/49.19	0.620	(0.43, 0.89)	0.0088	0.716	(0.49, 1.05)	0.089
COPD	76/138/55.07	0.790	(0.56, 1.11)	0.1761	0.986	(0.68, 1.43)	0.9428
ESRD	77/147/52.38	0.706	(0.51, 0.98)	0.0383	0.849	(0.59, 1.21)	0.3706
ALD	82/183/44.81	0.511	(0.38, 0.69)	<.0001	0.657	(0.47, 0.92)	0.0146
Medication
Metformin	423/763/55.44	0.768	(0.66, 0.90)	0.0012	0.757	(0.63, 0.92)	0.0043
Statin	152/281/54.09	0.750	(0.59, 0.96)	0.0207	0.805	(0.61, 1.06)	0.1186
Aspirin	224/423/52.96	0.704	(0.58, 0.86)	0.0007	0.868	(0.69, 1.09)	0.2312
Seeking behavior
Low (ref)	753/1354/55.61
Intermediate	863/1382/62.45	1.327	(1.14, 1.55)	0.0003	1.124	(0.95, 1.33)	0.1721
High	834/1306/63.86	1.410	(1.21, 1.65)	<.0001	1.293	(1.08, 1.54)	0.0045
SES
0–14,999	421/773/54.46
15,000–24,999	1376/2217/62.07	1.368	(1.16, 1.61)	0.0002	1.350	(1.13, 1.62)	0.0012
25,000–39,999	281/464/60.56	1.284	(1.02, 1.62)	0.0363	1.146	(0.89, 1.47)	0.2871
≥ 40,000	372/588/63.27	1.440	(1.16, 1.79)	0.0011	1.264	(1.00, 1.60)	0.0511
Urbanization
Urban (ref)	354/580/61.03
Suburban	1321/2198/60.10	0.962	(0.80, 1.16)	0.6824	0.994	(0.81, 1.22)	0.9553
Rural	775/1264/61.31	1.012	(0.83, 1.24)	0.9096	1.052	(0.83, 1.33)	0.6712
Hospital Level
Not center(ref)	1609/2679/60.06
Center	825/1323/62.36	1.102	(0.96, 1.26)	0.1612	1.043	(0.90, 1.21)	0.5777
No sonography	16/40/40.00	0.443	(0.23, 0.84)	0.0124	1.446	(0.71, 2.94)	0.3089
Hospital type
Private(ref)	1906/3153/60.06		3153			60.06
Public	528/849/62.36	1.076	(0.92, 1.26)	0.3565	1.116	(0.95, 1.32)	0.1971
No sonography	16/40/40.00	0.436	(0.23, 0.82)	0.0106	–	–	–

HCC Hepatocellular carcinoma, Unscreened No sonography surveillance within 24 months, OR Odds ratio

After adjustment, a clear dose response between surveillance interval and an early stage of HCC diagnosis remained: the 6-month group had the highest likelihood of being diagnosed with an early-stage HCC, followed by the 12-month group (OR = 0.69; 95% CI 0.55–0.85) and the 24-month group (OR = 0.36; 95% CI 0.27–0.47), the last by the unscreened group (OR = 0.30; 95% CI 0.22–0.40). Furthermore, patients with high outpatient utilization, and patients with a monthly wage 15,000–24,999 NTD had significantly more chance of early stage of HCC diagnosis. Past or current alcohol drinking, patients with ALD and metformin use were significantly associated with advanced stage of HCC diagnosis (Table 2).

Association between HCC surveillance and curative treatment

The proportions of patients in the 6-month, 12-month, 24-month and unscreened groups who received curative treatment were 36,17%, 28.08%, 22.25% and 19.73%, respectively. The univariate analyses showed that a clear dose response pattern was observed between surveillance interval and the chance of curative treatment. The 6-month group had the highest likelihood of patients who received the curative treatment, followed by the 12-month group (OR = 0.69; 95% CI 0.56–0.84) and the 24-month group (OR = 0.51; 95% CI 0.39–0.66), the last by the unscreened group (OR = 0.43; 95% CI 0.33–0.57). Likewise, patients with a monthly wage ≥ 40,000 NTD and patients primarily cared for by medical centers were more likely to be received with curative treatment. In contrast, patients with current alcohol drinking behavior, patients with mental illness disease and CVD diseases were significantly less likely to be received curative treatment (Table 3).

Table 3.

Univariable and multivariable analyses of variables associated with curative treatment of HCC

	Event/Patients/%	Univariate			Multivariable
		OR	95%	P	OR	95%	P
Screening interval
0–6 months (ref)	1397/3862/36.7
7–12 months	146/520/28.08	0.689	(0.56, 0.84)	0.0003	0.721	(0.58, 0.89)	0.0025
13–24 months	77/346/22.25	0.505	(0.39, 0.66)	<.0001	0.584	(0.44, 0.77)	0.0001
Unscreened	72/365/19.73	0.434	(0.33, 0.57)	<.0001	0.513	(0.38, 0.69)	<.0001
Gender
Male (ref)	1397/3862/36.17
Female	146/520/28.08	1.100	(0.98, 1.24)	0.1089	1.022	(0.89, 1.18)	0.7673
Age
Age < 55	138/417/33.09
55 ≦ Age < 65	478/1304/36.66	1.170	(0.93, 1.48)	0.1867	1.047	(0.81, 1.35)	0.726
65 ≦ Age < 75	625/1755/35.61	1.118	(0.89, 1.40)	0.3329	1.070	(0.83, 1.39)	0.6082
Age ≥ 75	451/1617/27.89	0.782	(0.62, 0.99)	0.0371	0.900	(0.68, 1.18)	0.4494
Alcohol drinking
Never	1034/2482/41.66
Past	171/431/39.68	0.921	(0.75, 1.14)	0.44	0.986	(0.76, 1.28)	0.9136
Current	143/401/35.66	0.776	(0.62, 0.97)	0.0236	0.889	(0.68, 1.16)	0.386
Missing	344/1779/19.34	0.336	(0.29, 0.39)	<.0001	0.962	(0.55, 1.69)	0.8937
Smoking
Never	985/2354/41.84
Past	51/125/40.80	0.958	(0.66, 1.38)	0.8177	1.016	(0.67, 1.54)	0.9398
Current	283/745/37.99	0.851	(0.72, 1.01)	0.0622	0.951	(0.76, 1.20)	0.6668
Missing	373/1869/19.96	0.347	(0.30, 0.40)	<.0001	0.775	(0.49, 1.24)	0.2868
Co-morbidities
DM	985/2354/41.84	0.896	(0.77, 1.04)	0.1522	0.983	(0.82, 1.18)	0.8497
Mental illness	51/125/40.80	0.426	(0.24, 0.75)	0.0028	0.450	(0.25, 0.81)	0.0075
CVD	283/745/37.99	0.610	(0.43, 0.87)	0.0069	0.735	(0.50, 1.08)	0.1132
COPD	373/1869/19.96	0.819	(0.60, 1.13)	0.2209	1.021	(0.72, 1.44)	0.9041
ESRD	985/2354/41.84	0.971	(0.71, 1.33)	0.8563	1.184	(0.84, 1.67)	0.3344
ALD	51/125/40.80	0.757	(0.56, 1.02)	0.0629	0.861	(0.62, 1.20)	0.3719
Medication
Metformin	300/940/31.91	0.930	(0.80, 1.08)	0.3461	0.837	(0.70, 1.00)	0.0548
Statin	124/334/37.13	1.202	(0.96, 1.51)	0.1176	1.174	(0.91, 1.52)	0.2207
Aspirin	164/513/31.97	0.939	(0.77, 1.14)	0.5251	0.937	(0.75, 1.17)	0.5669
Seeking behavior
Low (ref)	524/1695/30.91
Intermediate	596/1730/34.45	1.175	(1.02, 1.36)	0.0275	1.056	(0.90, 1.23)	0.4932
High	572/1668/34.29	1.166	(1.01, 1.35)	0.0367	1.097	(0.93, 1.29)	0.2647
SES
0–14,999	301/992/30.34
15,000–24,999	930/2801/33.20	1.141	(0.98, 1.33)	0.0984	1.133	(0.96, 1.35)	0.1529
25,000–39,999	186/581/32.01	1.081	(0.87, 1.35)	0.4891	0.937	(0.74, 1.19)	0.5912
≥ 40,000	275/719/38.25	1.422	(1.16, 1.74)	0.0007	1.284	(1.03, 1.60)	0.0241
Urbanization
Urban (ref)	258/723/35.68
Suburban	902/2747/32.84	0.881	(0.74, 1.05)	0.1487	0.973	(0.81, 1.17)	0.7754
Rural	532/1623/32.78	0.879	(0.73, 1.06)	0.1692	0.991	(0.80, 1.23)	0.9302
Hospital Level
Not center(ref)	1102/421/32.21
Center	581/1622/35.82	0.851	(0.75, 0.96)	0.0112	1.123	(0.98, 1.29)	0.0944
No sonography	9/50/18.00	0.393	(0.19, 0.82)	0.0121	0.876	(0.40, 1.94)	0.7444
Hospital type
Private(ref)	1318/3943/33.43
Public	365/1100/33.18	0.989	(0.86, 1.14)	0.8791	1.012	(0.87, 1.18)
No sonography	1318/3943/33.43	0.437	(0.21, 0.90)	0.0252	–	–	–

HCC Hepatocellular carcinoma, Unscreened No sonography surveillance within 24 months, OR Odds ratio

After adjustment, a clear dose response between surveillance interval and curative treatment of HCC remained: the 6-month group had the highest likelihood of being received curative therapy, followed by the 12-month group (OR = 0.72, 95% CI 0.58–0.89) and the 24-month group (OR = 0.58; 95% CI 0.44–0.77), the last by the unscreened group (OR = 0.51; 95% CI 0.38–0.69). Patients with mental illness disease and were significantly less likely to be received curative treatment. Patients with a monthly wage ≥ 40,000 NTD were more likely to be received with curative treatment (Table 3).

Association between HCC surveillance and all-cause mortality

In total, 71.53% (3643/5093) of the patients died during the follow-up period. The cumulative mortalities of the 6-month, 12-month, 24-month, and unscreened groups were 69.14%, 73.85%, 84.10%, and 81.64%, respectively. The univariate analyses showed that the 6-month group had the least likelihood of all-causes mortality, followed by the 12-month group (HR = 1.21; 95% CI 1.08–1.34), the 24-month group (HR = 1.74; 95% CI 1.55–1.97), and the unscreened group (HR = 1.71; 95% CI 1.52–1.93). Likewise, the odds of all-cause mortality were significantly reduced in female sex, patients with aged 55–64, and patients with a monthly wage 15,000–24999 NTD, 25,000–39,999, NTD, > 40,000 NTD. In contrast, patients aged 75 years or older, patients with past or current alcohol drinking behavior, patients with DM, mental illness, CVD, ESRD, COPD, ALD, and patients with metformin use, statin use, or aspirin use had a significantly higher chance of all-cause mortality. Similarly, patients living in suburban area, in rural area, and patients who primarily cared for at medical center also had significantly higher odds of all-cause mortality (Table 4).

Table 4.

Univariable and multivariable analyses of variables associated with all-mortality of HCC

	Event/Patients/%	Univariate			Multivariable
		HR	95%	P	HR	95%	P
Screening interval
0–6 months (ref)	2670/3862/69.14
7–12 months	384/520/73.85	1.206	(1.08, 1.34)	0.0006	1.134	(1.02, 1.26)	0.0223
13–24 months	291/346/84.10	1.744	(1.55, 1.97)	<.0001	1.570	(1.39, 1.77)	<.0001
Unscreened	298/365/81.64	1.708	(1.52, 1.93)	<.0001	1.520	(1.33, 1.73)	<.0001
Gender
Male (ref)	1798/2459/73.12
Female	1845/2634/70.05	0.900	(0.84, 0.96)	0.0015	0.951	(0.88, 1.02)	0.181
Age
Age < 55	287/417/68.82
55 ≦ Age < 65	814/1304/62.42	0.824	(0.72, 0.94)	0.0047	0.959	(0.83, 1.10)	0.5547
65 ≦ Age < 75	1231/1755/70.14	1.004	(0.88, 1.14)	0.9516	1.098	(0.95, 1.26)	0.1927
Age ≥ 75	1311/1617/81.08	1.479	(1.30, 1.68)	<.0001	1.373	(1.19, 1.58)	<.0001
Alcohol drinking
Never	1577/2482/63.54
Past	294/431/68.21	1.154	(1.02, 1.31)	0.0238	1.043	(0.90, 1.21)	0.5903
Current	289/401/72.07	1.237	(1.09, 1.40)	0.0009	1.043	(0.90, 1.21)	0.5835
Missing	1483/1779/83.36	1.444	(1.35, 1.55)	<.0001	0.918	(0.68, 1.24)	0.578
Smoking
Never	1491/2354/63.34
Past	82/125/65.60	1.173	(0.94, 1.47)	0.1605	1.105	(0.87, 1.41)	0.415
Current	522/745/70.07	1.227	(1.11, 1.36)	<.0001	1.099	(0.96, 1.25)	0.16
Missing	1548/1869/82.83	1.454	(1.35, 1.56)	<.0001	1.041	(0.80, 1.35)	0.7629
Co-morbidities
DM	768/972/79.01	1.316	(1.22, 1.43)	<.0001	1.116	(1.02, 1.23)	0.0217
Mental illness	75/85/88.24	1.475	(1.17, 1.85)	0.0009	1.327	(1.05, 1.68)	0.0182
CVD	144/170/84.71	1.671	(1.42, 1.98)	<.0001	1.322	(1.12, 1.57)	0.0013
COPD	161/189/85.19	1.653	(1.41, 1.94)	<.0001	1.244	(1.06, 1.47)	0.0091
ESRD	148/181/81.77	1.380	(1.17, 1.63)	0.0001	0.982	(0.83, 1.17)	0.8326
ALD	199/232/85.78	1.623	(1.41, 1.87)	<.0001	1.451	(1.24, 1.70)	<.0001
Medication
Metformin	717/940/76.28	1.289	(1.19, 1.40)	<.0001	1.184	(1.07, 1.31)	0.0007
Statin	245/334/73.35	1.306	(1.15, 1.49)	<.0001	1.151	(1.00, 1.32)	0.0483
Aspirin	399/513/77.78	1.565	(1.41, 1.74)	<.0001	1.331	(1.19, 1.49)	<.0001
Seeking behavior
Low (ref)	1202/1695/70.91
Intermediate	1223/1730/70.69	0.963	(0.89, 1.04)	0.3514	1.010	(0.93, 1.10)	0.8204
High	1218/1668/73.02	1.019	(0.94, 1.10)	0.6389	0.976	(0.90, 1.06)	0.5842
SES
0–14,999	757/992/76.31
15,000–24,999	2023/2801/72.22	0.893	(0.82, 0.97)	0.0078	0.879	(0.81, 0.96)	0.0039
25,000–39,999	395/581/67.99	0.791	(0.70, 0.89)	0.0002	0.902	(0.80, 1.02)	0.1021
> 40,000	468/719/65.09	0.738	(0.66, 0.83)	<.0001	0.855	(0.76, 0.96)	0.0088
Urbanization
Urban (ref)	482/723/66.67
Suburban	1968/2747/71.64	1.142	(1.03, 1.26)	0.009	1.046	(0.94, 1.16)	0.3869
Rural	1193/1623/73.51	1.217	(1.09, 1.35)	0.0003	1.041	(0.93, 1.17)	0.5034
Hospital Level
Not center(ref)	2509/3421/73.34
Center	1091/1622/67.26	1.233	(1.15, 1.32)	<.0001	0.833	(0.77, 0.90)	<.0001
No sonography	2509/3421/73.34	1.658	(1.22, 2.25)	0.0011	0.791	(0.57, 1.10)	0.1602
Hospital type
Private(ref)	2817/3943/71.44
Public	783/1100/71.18	0.999	(0.92, 1.08)	0.9901	0.984	(0.91, 1.07)	0.6963
No sonography	43/50/86.00	1.440	(1.07, 1.95)	0.0175	–	–	–

HCC Hepatocellular carcinoma, Unscreened No sonography surveillance within 24 months, HR Hazard ratio

After adjustment, the 6-month group had the least likelihood of all-cause mortality, followed by the 12-month group (HR = 1.13; 95% CI 1.02–1.26), the 24-month group (HR = 1.57; 95% CI 1.39–1.77), and the unscreened group (HR = 1.52; 95% CI 1.33–1.73). Likewise, patients with a monthly wage15000-24999 NTD and more than 40,000 NTD, and patients who primary cared for at medical center were less likelihood of all-cause mortality. In contrast, patients with aged more than 75, patients with DM, mental illness, CVD, ESRD, and ALD, and patients with metformin use, or aspirin use were significantly more likely to be associated with all-cause mortality (Table 4).

Even after adjusting for lead-time bias, the 6-month group had the least likelihood of all-cause mortality, followed by the 12-month group (HR = 1.21; 95% CI 1.09–1.34, Mode 1; HR = 1.21; 95% CI 1.09–1.34, Mode 2; HR = 1.13; 95% CI 1.02–1.26, Mode 3), the 24-month group (HR = 1.75; 95% CI 1.55–1.98, Mode 1; HR = 1.76; 95% CI 1.55–1.98, Mode 2; HR = 1.50; 95% CI 1.33–1.69, Mode 3), and the unscreened group (HR = 1.54; 95% CI 1.37–1.94, Mode 1; HR = 1.43; 95% CI 1.27–1.61, Mode 2; HR = 1.33; 95% CI 1.18–1.50, Mode 3) (Table 5). Compared with 12-month group, the impact of prognostic factors on survival at the time of HCC diagnosis were shown below. Cirrhotic HCV patients with an AFP less than 20 ng/ml (P = 0.0206), with a MELD score less than 20 (P = 0.0134), with cirrhosis duration between 3–5 years (P = 0.0037) had the better survival rate in the 6-month group (Figs. 2, 3, and 4).

Table 5.

Association between surveillance intervals and all-cause mortality after adjusting for lead-time bias

		Model 1
Surveillance time	Sojour time (days)	Median survival (years)	HR (95% CI)
0–6 months	70	2.23	1
7–12 months	70	1.84	1.21 (1.09, 1.34)
13–24 months	70	1.22	1.75 (1.55, 1.98)
unscreened groups	None	1.34	1.54 (1.37, 1.94)
		Model 2
Surveillance time	Sojour time (days)	Median survival (years)	HR (95% CI)
0–6 months	140	2.04	1
7–12 months	140	1.65	1.21 (1.09, 1.34)
13–24 months	140	1.03	1.76 (1.55, 1.98)
unscreened groups	None	1.34	1.43 (1.27. 1.61)
Surveillance time	Sojour time (days)	Median survival (years)	HR (95% CI)
		Model 3
0–6 months	220	1.83	1
7–12 months	140	1.65	1.13 (1.02, 1.26)
13–24 months	70	1.22	1.50 (1.33, 1.69)
unscreened groups	None	1.34	1.33 (1.18, 1.50)

Model 1: Lead time: 70 days

Model 2: Lead time: 140 days

Model 3: Lead time: (220 days, 140 days, 70 days)

HR Hazard ratio

Fig. 2.

Comparison of the impact of tumor biology on the survival curve between 6-month and 12-month (Blue line: 6-month, Red line: 12-month). Tumor size ≤ 3 cm (a), Tumor size > 3 cm (b), AFP level < 20 (P = 0.0206) (c), AFP level: 20–40 (Fig. 2 d), AFP level > 400 (e)

Fig. 3.

Comparison of the impact of liver function on the survival curve between 6-month and 12-month (Blue line: 6-month, Red line: 12-month). MELD score < 20 (P = 0.0134) (a), MELD score ≥ 20 (b), Child–Pugh score A (c), Child–Pugh score B (d), Child–Pugh score C (e)

Fig. 4.

Comparison of the impact of cirrhosis duration on the survival curve between 6-month and 12-month (Blue line: 6-month, Red line: 12-month). Liver cirrhosis duration < 3 years (a), liver cirrhosis duration 3–5 years (P = 0.0037) (b), liver cirrhosis duration > 5 years (c)

Discussion

For years, hepatologists in Taiwan have consistently advocate regular ultrasonography screening for high-risk populations. Compared with the HCC surveillance pattern between 2002 and 2007 reported in the literature [11], our results showed that the HCC surveillance rate among HCV-infected individuals with cirrhosis in Taiwan had improved between 2008 and 2016. About 14% of HCV cirrhotic patients, one type of high-risk subpopulations, did not undergo an ultrasonography screening every 6 months or one year. Patients with HCV-related cirrhosis had an estimated annual risk of developing HCC of 3.5% [25].

The optimal interval for screening remains a challenge, particularly regarding the effectiveness of 6-month versus 12-month surveillance, including the diagnosis of early-stage HCC and overall mortality [11, 26]. Our results show that, for HCV-infected patients with cirrhosis, performing ultrasound examinations at 6-month intervals had a better chance of detecting early-stage HCC and led to better overall survival compared with 12-month intervals. Observational studies are susceptible to lead time bias. Thus, our study accounted for the lead time bias by conducting sensitivity analysis based on different sojourn time (70 days, 140 days to 220 days) to strengthen the reliability of our observations. This approach may not completely eliminate, but at least attenuate, the potential lead time bias of observational studies. Even after adjusting for the HCC sojourn time of 220 days, the adjusted survival rates of 6-month HCC surveillance remained to be statistically significantly better than that for 12-month HCC surveillance. The lead time bias depends on the growth rates of the HCC and the surveillance interval. According to previous studies, the average HCC tumor doubling time (TDT) were 4–5 months [27, 28]. In addition, a previous study reported that after 10 years of follow-up in 1380 patients with Child–Pugh score A or B, the median lead-time was 7.2 months with semiannual surveillance and 4.1 months with annual surveillance [29]. Therefore, we used a variety of sojourn times ranging between 70 and 220 days to address possible lead-time bias. Although we could not fully eliminate the bias resulting from possible early diagnosis of preclinical HCC caused by 6-month surveillance, this approach could help to avoid overestimating the surveillance benefit impact on survival rates.

Our findings also illustrate several risk factors associated with advanced HCC diagnosis and higher mortality. Our study showed that patients with drinking habits and ALD had a higher likelihood of being diagnosed with more advanced HCC. Additionally, patients with ALD have lower overall survival rates. One plausible explanation may be that patients with ALD have a greater chance of developing hepatic steatosis or steatohepatitis, and then developing cirrhosis or HCC [30]. Furthermore, patients with both ALD and HCV infection may experience synergistic disease progression, accelerating the development of cirrhosis and significantly increasing the risk of HCC. According to a recent meta-analysis study, the combined effect of HCV and alcohol leads to a 42-fold increase in the incidence of HCC [31]. Such patients may require more frequent HCC surveillance due to the synergistic effects of both diseases on oxidative stress, immune components, and carcinogenic mechanism [32].

Other predicted factors associated with increased overall mortality included patients older than 75 years, diabetes, mental illness, CVD, and ESRD. Furthermore, in our study, patients taking aspirin and metformin were also found to have poorer survival rates. In the United States, all-cause mortality is higher in patients with HCC who are 70 years of age or older [33]. In compensated cirrhosis, complications and mortality are increased when patients coexist with DM [34]. Even in the case of DAA cure, DM remains a risk factor for liver-related mortality [35], and good glycemic control is an important factor in reducing liver-related mortality [34, 36]. Compared with the general population, patients with mental illness have a higher prevalence of chronic comorbidities. Cancer and liver cirrhosis are two major chronic diseases that increase the risk of all-cause mortality in people with mental illness compared with those without mental illness [37]. In past studies, aspirin and metformin have been found to reduce liver cancer incidence, liver-related mortality, and all-cause mortality [37–39]. But patients with comorbidities such as DM or CVD had higher baseline mortality, attenuating the survival benefit of aspirin or metformin. Our study shows that the protective effect of drugs on survival is relatively small compared with the risk of death from comorbidities.

Patients with more frequent outpatient visits were more likely to be diagnosed with HCC at an earlier stage. We found that SES and urbanization did not affect HCC screening compliance. Higher SES patients including patients with a monthly wage of 15,000–24999 NTD and ≥ 40,000 NTD had better survival. Urbanization level did not affect the HCC early diagnosis and survival rate. Moreover, patients those who received care primarily in medical centers had had better survival. It is likely that these patients who may have paid more attention to the progression of liver disease, have better adherence to surveillance guidelines, and lead to better outcomes.

The duration and severity of cirrhosis may significantly influence a patient's willingness to undergo HCC surveillance. Therefore, we used Taiwan cancer registry data to analyze the impact of potential prognostic factors at the time of HCC diagnosis, including tumor size, AFP level, MELD score, Child score, and duration of cirrhosis, on survival rates. Our study found that HCV cirrhosis patients with AFP values below 20 ng/ml or MELD scores below 20 at the time of HCC diagnosis had better survival in the 6 M group compared with the 12-month group. Shorter surveillance intervals (6-month) appear to result in better survival in this subpopulation. In addition, the same results occurred in a subpopulation of patients with cirrhosis duration of 3–5 years, with shorter surveillance intervals (6-month) having better survival rates. Past studies have shown that patients with cirrhosis may initially comply with regular HCC surveillance but gradually neglect and reduce HCC monitoring over the next 2 to 4 years [18]. We included HCC cases up to March 2017, so cases with at least 2 years of follow-up from the index date if no deaths occurred. In our study population, more than 60% had cirrhosis lasting more than 3 years. Moreover, we found that in patients with tumor size ≤ 3 cm, Child–Pugh score A, and cirrhosis duration > 5 years, the survival curve of 6-month was better than that of 12-month throughout the follow-up, but it did not reach the level of statistical significance. The statistically insignificant findings could likely due to small sample sizes. Therefore, surveillance intervals of these subgroups shall be further investigated with larger sample sizes and if the differences in clinical benefits between 6-month and 12-month intervals are proven not significant in these subgroups using larger sample sizes, clinically, there will be no need for semi-annual HCC surveillance.

According to current guidelines, HCC surveillance should be conducted in all patients with cirrhosis, regardless of etiology, as they are considered at increased risk for HCC [40]. In patients with ALD cirrhosis, however, surveillance efforts are often hampered by social deprivation and limited access to healthcare resources [41]. Nevertheless, when HCC is detected through regular surveillance in patients with ALD, the rate of receipt of curative treatment is high and comparable to that in patients with viral hepatitis–related HCC [41]. Notably, a significant proportion of HCC cases associated with metabolic dysfunction–associated steatotic liver disease (MASLD) occur in the absence of cirrhosis [42]. Moreover, MASLD and chronic HBV infection act synergistically to accelerate liver disease progression [43]. In contrast, fatty liver without metabolic dysfunction does not appear to be associated with adverse clinical outcomes in patients with HBV infection [44]. In patients with HCV infection, systemic metabolic alterations and hepatic steatosis contribute to HCC development and progression [45]. Therefore, our findings on optimal surveillance intervals might be more applicable to patients with HBV- and alcohol-related cirrhosis. In patients with MASLD, further validation is warranted using real-world data, with adjustments for lifestyle factors, comorbidities, and interactions with viral hepatitis—including the impact of antiviral therapy—especially given the evolving landscape of HCC in the post–direct-acting antiviral (DAA) era.

We acknowledge that this study has some limitations. First, we first confirmed the diagnosis of HCC and then evaluated the impact of surveillance intervals on early diagnosis and survival of HCC in this study. Identifying HCC surveillance intervals for all patient is difficult because in the real world, patients often do not strictly adhere to HCC surveillance intervals, and actual surveillance intervals may vary at different time points. We therefore used the interval between the last HCC surveillance date and the index date (i.e. 90 days before the date of HCC diagnosis) to define the HCC surveillance interval. Second, due to the nature of observational studies, confounding by indication may be another source of bias. Although we have included many potential confounders in our analyses, due to data limitations of administrative and claims data, potential important confounders such as neutrophil-to-lymphocyte ratio (NLR) and Eastern Cooperative Oncology Group (ECOG) [46, 47] are not included and might lead to confounding biases. Third, since we analyzed the NHI data, those HCC surveillance screening paid out-of-pocket by patients were not included in our analyses. Fourth, as 36.3% of alcohol intake, 36.30% of smoking, 39.31% of AFP level, 41.02% of Child–Pugh score, and 29.85% of MELD score data were missing, we conducted sensitivity analyses by excluding individuals with missing values. The results remained robust. The missing data of five variables did not seem to significantly affect our findings. Fifth, due to the data limitation, our finding cannot evaluate the influence of the introduction of DAA on survival benefits of different HCC surveillance intervals. Taiwan has started to cover DAA since 2018. Further research with more recent data will help to explore the effectiveness of semi-annual surveillance and surveillance strategies before and after the introduction of DAA. We will analyze the effects of different surveillance intervals on survival in patients with index dates before and after 2018, and further compare the survival differences between patients who used DAA after 2018 and those who did not. Finally, whether our results can be extrapolated to Western populations with patients with cirrhotic HCV infection will require further analyses.

This research has several strengths as follows. First, our study used a national population-based study, which had the advantage of a sufficiently large sample size that allowed us to control for many potential confounders, such as patient comorbidities and medications used. Second, the clinical progression of HCC varies depending on the etiology. Thus, we utilized real world data to only analyze the impact of HCC surveillance intervals on early diagnosis and survival rates in patients with HCV-related cirrhosis, which can provide precise surveillance guidelines for patients with HCV-related cirrhosis. Third, we used data from the Taiwan Cancer Registry in our analysis, which provides pathology reports and allows us more accurately identify HCC events and HCC staging. Also, the data allows us to probe how a patient's liver function influences survival outcomes. Fourth, we also attempted to adjust for lead time bias to enhance validity of our findings.

In conclusion, our results show that a 6-month surveillance interval can significantly improve the detection rate of early-stage HCC, curative treatment, and prolong the overall survival of cirrhotic HCV patients. More specifically, compared with the 12-months interval surveillance, patients with cirrhotic HCV infection who had an AFP value below 20 ng/ml, a MELD score below 20, or a cirrhosis duration of 3–5 years at the time of HCC development, 6-months interval surveillance would also lead to better survival. Therefore, these patient subgroups, which are generally considered to have a better prognosis, should also receive HCC surveillance every 6 months.

Abbreviations

HCC

Hepatocellular carcinoma

BCLC

Barcelona Clinic Liver Cancer

HBV

Hepatic B virus

HCV

Hepatic C virus

AASLD

American Association for the Study of Liver Disease

EASL

European Association for the Study of the Liver

NHIRD

National Health Insurance Research Database

Mcohort

Month cohort

CHB

Chronic hepatitis B

DM

Diabetes mellitus

CAD

Coronary artery disease

CVD

Cerebrovascular disease

COPD

Chronic obstructive pulmonary disease

ESRD

End stage renal disease

ALD

Alcohol-related liver diseases

TACE

Trans-arterial chemoembolization

SES

Socioeconomic status

NT$

New Taiwan dollars

NSAIDs

Non-steroidal anti-inflammatory drugs

AFP

Alpha-Fetoprotein

MELD

Model for End-Stage Liver Disease

OR

Odds ratios

HR

Hazards ratios

MAR

Missing at random

TDT

Tumor doubling time

DAA

Direct-acting antiviral

NLR

Neutrophil-to-lymphocyte ratio

ECOG

Eastern Cooperative Oncology Group

MASLD

Metabolic dysfunction-associated steatotic liver disease

Authors’ contributions

Performing database management, data analysis and statistic testing—Yu-Chin Chen and Hsiao-Yun Hu. Study design and implementation of the study—Shen-Shong Chang and Nicole Huang. Critical revision—Nicole Huang and Yung-Feng Yen. Writing—Shen-Shong Chang and Nicole Huang.

Funding

There is no financial support.

Data availability

Availability of data and materials Data and materials are available from the Collaboration Center of Health Information Application of the Ministry of Health and Welfare. The data used in this study are available upon application from the Health and Welfare Data Science Center. The data cannot be made publicly available in paper, supplementary files, or a public repository due to compliance with the Personal Information Protection Act in Taiwan. Requests for the data can be sent directly to the Health and Welfare Data Science Center (https://dep.mohw.gov.tw/DOS/cp-5119–59,201-113.html).”

Declarations

Ethics approval and consent to participate

We confirmed that all methods were performed in accordance to the Declaration of Helsinki. To assure privacy and confidentiality, all data analyses were carried out onsite within the Collaboration Center of Health Information Application of the Ministry of Health and Welfare. We abided all government regulations in data management and analyses. This study was approved by The Institutional Review Board (IRB) of Taipei City Hospital approved this study (TCHIRB-11110004-E), and the need for informed consent was waived.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.